edfas.org 7 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 27 NO. 4 when the frequency was set to 5 kHz. Strong signals are observed along the outline of the pattern. This is unnecessary noise (called scan noise) because it is not a signal of thermo-reflectance. In the case of 670 nm, the scan noise level is high, in the range of 1 to 10 kHz. Above 10 kHz, the scan noise level drops to the floor level of the light source. Regarding the detection limit of the TD Imaging system, noise equivalent temperature (NET) is of significant interest. ∆Pr is converted to photocurrent I by an avalanche photodiode (APD), then Eq 7 is rewritten as: (Eq 8) NET depends on the system, the sample surface material, its structure, and condition of the surface. Another general parameter to characterize the system would be the relative noise equivalent photocurrent (RNEP), which is calculated by dividing noise equivalent photocurrent ∆I by DC photocurrent I flowing through the APD. RNEP is ultimately limited by relative intensity noise (RIN) of light source. Table 3 shows the RNEP and NET in 670 nm and 1300 nm HILs. This is calculated by the TR Coeff in Table 2, and ω is 100 kHz. PROOF-OF-CONCEPT EXPERIMENT To test the capability of TD Imaging, a simple test structure fabricated with an 8 μm-wide Al wire (resistance = and ω is 100 kHz. PROOF-OF-CONCEPT EXPERIMENT To test the capability of TD Imaging, a simple test structure fabricated with an 8 μm- wide Al wire (resistance = 5.6 ohm) is shown in Fig. 3a. Significant Joule heating is expected at this wire. For modulating the heat, a square wave bias with a voltage swing of 0 to 210 mV at certain frequencies was applied to the sample. This sample was analyzed with the TD Imaging tool, first with the 1300 nm, then with 670 nm light sources. Modulation frequencies were set at 8 kHz. A Mitutoyo objective lens of 20× with a numerical aperture (NA) of 0.40, is transparent enough to see the two wavelengths. Multiple probing beam powers of no higher than 1.0 mW were used on the sample. TD Imaging in phase images with 670 nm and 1300 nm light source at probing beam power of 1.0 mW is shown in Figs. 3b and c, respectively, where pixel color indicates amount of perturbation in the reflected beam power or ∆Pr. To quantitatively compare the results among the techniques, the team evaluated signal-to-noise ratio (SNR). Figure 4 compares SNR of TD Imaging results with 1300 nm and 670 nm as a function of probing beam power. The dotted red circle symbol and dotted blue rectangular symbol shows data points with 670 nm and 1300 nm, respectively. For both light sources, a region is reached Fig. 2 Scan noise of the optics chart. (a) Pattern image. (b) TD Imaging amplitude. Magnification is 20×. The light source is 670 nm HIL. (a) (b) Table 3 Comparison of RNEP and NET of Al and Cu at two wavelengths in this system Wavelength of light source RNEP, db/Hz NET, mk/√Hz Cu Al 670 nm -135 8 3 1300 nm -145 5 5 Fig. 3 Pattern and TD Imaging in phase images of the test sample. (a) Infrared laser reflection image. The magnification is 20× × 4× zoom. 8-μm-wide Al wire is deposited on Si/SiO2 substrate. (b) and (c) TD Imaging in phase images with 670 nm and 1300 nm HILs. (a) (b) (c)
RkJQdWJsaXNoZXIy MTYyMzk3NQ==